Markus Rother

Vibed >50k lines of code in the last 18 hours. And stuff just works. now: 90% of time spent on design choices. then: 90% of time spent reading and writing code. Can't believe how there are still devs out there who don't get how much of a game changer this is.

Developers, try harder! There is a race going on, and us developers are losing. Product managers are experimenting with low code, producing prototypes, validating ideas with application generators. And us developers are sitting on our hands, thinking we will remain irreplacable forever. ❌ "Real use cases are too complex" ❌ "Where's the CI/CD pipeline?" ❌ "This will never achieve maturity" ❌ "And, lol, look how ChatGPT can't even count correctly" If that is you, you better make up your mind. AI will change coding forever. In particular, it will disrupt or has already disrupted how ideas are translated into functionality — which is what programming is about. Yes, I know you are using Copilot for code completion, because we are lazy and always have been. Is "using the tools" really enough? Own it! This could be the rebirth of verifiable software, yet nobody is talking about it. Or what about property based testing to force the spec to do what it promises? Merely mapping requirements to behaviour won't do much longer.

I often hear the analogy that #VibeCoding to generated high level code is like written high level code is to machine language. That is wrong! Compiling is a reduction. Information is lost in the process, such as meaningful variable or class names, or comments. When turning spec to code, on the other hand, a lot of information is added, such as naming conventions, choice of frameworks, etc. I am not opposing the beneficial hallucination. Not at all. All I am saying is that intent needs interpretation, and interpretation is dependent on the culture captured in the LLM. It is a nondeterministic process, inherently fragile.

nostr:nprofile1qqsdxpfv503a2ga3ajqxw843hws9z7302ghpj4mcmjpa6qagmp9pwrs8222jg, I like that you are bullish on #SpillmanChannels A few random thoughts: If the opener (sender) of a Spillman Channel has no way to recover funds other than locktime expiry, the barrier to entry is somewhat high. On the other hand, other recovery mechanisms probably require punishment and/or watchtowers of some sort. The risk of permanently locked and unused funds in a channel grows with the envisioned duration, so the incentive works against the problem to be solved (negative feedback loop). Therefore, there still is some trust involved. Then again, some trust can be assumed anywasy, if the sender is paying money to another party... What do you think?

There are only two ways to validate a Spec Driven Development pipeline: 1. Spec → Code → Spec :: no semantic loss 2. Code → Spec → Code :: no functional loss #vibecoding #SDD

Don’t worry about being wrong — that’s the point. Pair programming sessions don’t usually derail because of personality or skill gaps. They derail due to the lack of declared intent. ▸ "Let’s have a look." ▸ "I’m trying to understand." ▸ "It’s probably something very obvious." It sounds curious, humble, even open-minded — and yet, it usually means: ⚡ "I don’t know WTF I’m doing." ⚡ "I’m not ready to commit to an explanation." ⚡ "I’m hoping something will jump out at me." Without intent, without a hypothesis, every observation feels relevant. Nothing can be wrong — because nothing specific was expected. Obviously, not knowing is a valid starting point — maybe the only valid starting point in any endeavor, for that matter. However, declaring ignorance is a comfort trap. It can be a justification for remaining indecisive. Conversely, even the most ridiculous hypothesis shifts focus from activity to validation. You will be held accountable not only for your ridiculous thoughts but for the progress as well. So that's nice. That is the feedback loop you are entering when stating expectations. 💬 “I expect the result to be X.” 💬 ”If this condition is true, we should skip the database query.” 💬 “If X is null, Y should throw.” Say it aloud, even when not pair programming to turn wandering into a controlled experiment. And also, my self-talk signals to my spouse and daughter that I am in a state of deep concentration — rather not to be disturbed. Hashtag#PairProgramming Hashtag#SoftwareEngineering Hashtag#Agile